Computer program for evaluation, physicochemical characterization and optimization of competitive protein binding assays: comparison of four curve-fitting models in peptide and steroid radioimmunoassays.

A modular construced computer program for evaluation, physicochemical characterization and optimization of compititive protein binding assays is put forward. The program allows standard curve fitting by the following models: spline approximation, semi-logarithmic interpolation, logit-log regression and logit-log interpolation. The suitability of these models was studied in a series of routine steroid radioimmunoassays (deoxycorticosterone, deoxycortisol, cortisol and aldosterone) and peptide radioimmunoassays (angiotensin I and angiotensin II). The spline approximation model was used as an approximate reference model. In all assays the logit-log interpolation model exhibited the best correlation with the spline approximation model. In the steroid assays the logit-log regression model was found to be inferior to the other models; in the peptide assays it was equal to or superior to the semi-logarithmic interpolation model. The study on the physicochemical parameters of the antigen-antibody reactions demonstrated that affinity of steroid antibodies was much lower than that of the peptide antibodies. Consequently the degree of antibody saturation at zero dose in the routine peptide assays was greater than 50% and in the routine steroid assays was less than 50%. The relationship of assay characteristics to the applicability of particular curve-fitting models is discussed.